Ongoing Projects
|
Facilitating History Learning
|
|
Trust-based Personal
|
|
Collective Decision
Making
|
 |
Human-Centered
|
Trust-based Personal
Recommendation Manager
Main Ideas:
Record the recommendations you get: Recommenders do
Trust based recommender: If you trust Y's recommendations and Y trusts Z's recommendations, you may use Z's recommendations before Y's approval. How much you can trust Z, can be inferred using transitive trust inference systems such as GePuTTIS.
Don't forget the recommendations: Using WAP Interface + Web Interface, the service can be available anywhere.
Integration with Amazon & Bibsonomy
People Involved:
Related Publications:
Syavash Nobarany, Mona Haraty, Dan Cosley: GePuTTIS: General Purpose Transitive Trust Inference System For Social Networks, To appear in proceeding of AAAI Spring Symposium 2008 - Social Information Processing (U.S.A., California)
Human-centered Video Compression
The amount of visual information reaching human visual system (HVS) at any time is huge compared to what it can process. HVS handles this problem by choosing only a manageable size of information at a time. This way, information is processed partly “sequentially” and the amount of data that should be analyzed is reduced to manageable size. Visual attention is responsible for selection what data and in which order should be processed.
In this project, we will inspect human visual attention system and the ways that it can be used to provide more efficient compression for H.264 standard. The process includes 1. Attention-based ROI detection and 2. Adaptive quantization and macroblock ordering.
Current state: Evaluation
Under development paper:
Mona Haraty, Syavash Nobarany, Mahmoud
Reza Hashemi:
Human-centered Video Compression on H.264 Based on a Vector-Space Model
People Involved:
Facilitating History Learning Using
a Zoomable Adaptive Timeline
Description:
Learning history by reading textual information is not efficient. An advanced timeline is proposed and implemented for visualizing time-based data. The components of the timeline, which are year, month, and day are zoomable. Color-coding is used to provide basic information about years by looking at the timeline at a glance. The software is implemented using Piccolo, an information visualization framework from University of Maryland.
Key Features:
Screen Shots:
 |
Wide View:
|
 |
Search View :
|
 |
Zoom View
|
People Involved:
Related Publications:
Mona Haraty, Syavash Nobarany, Dan Cosley, Ahmad Khonsari: Facilitating history learning using a zoomable adaptive timeline, (Under preparation)
Collective Decision Making
Based on Trust Relationships
Description:
In traditional communities, a moderator is chosen for future decisions of the group. The moderator is usually chosen by previous moderators of the group that is called an autocratic system. In other communities that are using traditional democracy, important decisions such as choosing moderators are made through general polls.
In democratic groups, where all people are equal and decisions are supposed to be in favor of all members, general polls and elections are the best known decision making system. The traditional election system has many known shortcomings. For example, consider an election in a democratic country. Of course, in no election all people attend. It means that the result of the election does not indicate vote of every member. Another problem in traditional elections has come from shortcomings of simple democracy. In traditional democracy, vote of any member of a group with any level of merit and abilities, has the same value in the election. In regular elections any person who has the age more than a threshold can vote. This is because there is no simple way to determine merit of all voters. In online communities the problem is somehow different, because we can expect people to express their trust to other members, so these expressions can be used when they do not want or cannot attend (vote) in an election or poll.
People Involved:
